在運動風氣盛行的情況下,各地的運動彩券風潮也非常興盛。而台灣為了提升運動風氣,以及提供基層運動經費也由富邦銀行代理發行了富邦運動彩券。在運動彩券的投注類型日益複雜,以及投注人口和投注金額的增加,運動項目賽果相關的分析也相對熱門。過去運動彩券相關論文往往著重在球賽數據的分析,而較少針對投注交易資料做分析。考量莊家為了能夠將交易風險降到最低,本篇論文針對投注資料使用不同的分類模型做預測,以期能讓莊家在比賽結果出現前能夠做出適當的風險控管。本研究使用三種常見的模型做預測比較:決策樹、支援向量機、類神經網路,預測出各筆交易可能的盈虧情形,並比較決策樹、支援向量機及類神經網路此三種預測方式。其中決策樹法能夠明顯增加莊家的盈餘,在資料集6可以達到準確率67.3%及每筆投注莊家賺取778.52元的結果。未來希望能夠針對資料集挑選出適合的SVM模型內核、找出更好的績效評估方法並且將訓練出來的模型建製成及時的決策支援系統。 Latest research on sport lottery mostly focus on the analysis of game's statistics, but less on the way how people bet .To lower down the cost of the bookmaker, we use different classification models for predicting the game's result. The models can also help bookmakers control their risk before the end of the game and get profit. In this research , we use three popular and easy-to-use models : decision-tree、support-vector-machine、neuron-network. We predict the result of each sport lottery ticket, and we find out that the decision-tree model will increase the profit of the bookmaker evidently. In data set 6,the accuracy is 67.53 % and bookmaker will earn 778.52 NT dollars each bet. For our future improvement, we can adjust the SVM kernel to each data sets、establish the performance evaluation system and build up a real-time decision support system for bookmakers.